fix(server): eliminate ~2-min test startup delay on Windows (#122)

* fix(server): eliminate ~2-min test startup delay on Windows

Replace `npx playwright` with a direct `node @playwright/test/cli.js`
spawn in both the /api/playwright/run and /api/run-dynamic-test handlers.
npx + cmd.exe shell resolution was adding 30–90 s of overhead before any
test code ran on Windows. Also fire the Render warmup ping concurrently
with Playwright startup (was sequential) and cap its timeout at 30 s
(was 55 s), removing a second source of blocking delay.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* style: fix prettier formatting in CHANGELOG, README, wiki Home

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: MyNameIsEdi

.github/wiki/Home.md

@@ -8,14 +8,14 @@ For the full documentation see the [README](https://github.com/MyNameIsEdi/open-
 
 ## Quick links
 
-| Topic | Description |
-|---|---|
-| [Quick Start](https://github.com/MyNameIsEdi/open-qa#-quick-start) | Clone, install, configure LLM, run |
-| [Playwright Dashboard](https://github.com/MyNameIsEdi/open-qa#-features) | Run tests, stream live output, view KPIs |
-| [CLI Agents](https://github.com/MyNameIsEdi/open-qa#-cli-agents) | Self-healing, Auto-POM, Bug Triage, A11y, Visual Regression, Data Gen |
-| [Switching LLMs](https://github.com/MyNameIsEdi/open-qa#-switching-llm-providers) | Gemini ↔ Ollama toggle |
-| [Contributing](https://github.com/MyNameIsEdi/open-qa/blob/main/CONTRIBUTING.md) | PR guidelines, code style, branch naming |
-| [Changelog](https://github.com/MyNameIsEdi/open-qa/blob/main/CHANGELOG.md) | Release notes |
+| Topic                                                                             | Description                                                           |
+| --------------------------------------------------------------------------------- | --------------------------------------------------------------------- |
+| [Quick Start](https://github.com/MyNameIsEdi/open-qa#-quick-start)                | Clone, install, configure LLM, run                                    |
+| [Playwright Dashboard](https://github.com/MyNameIsEdi/open-qa#-features)          | Run tests, stream live output, view KPIs                              |
+| [CLI Agents](https://github.com/MyNameIsEdi/open-qa#-cli-agents)                  | Self-healing, Auto-POM, Bug Triage, A11y, Visual Regression, Data Gen |
+| [Switching LLMs](https://github.com/MyNameIsEdi/open-qa#-switching-llm-providers) | Gemini ↔ Ollama toggle                                                |
+| [Contributing](https://github.com/MyNameIsEdi/open-qa/blob/main/CONTRIBUTING.md)  | PR guidelines, code style, branch naming                              |
+| [Changelog](https://github.com/MyNameIsEdi/open-qa/blob/main/CHANGELOG.md)        | Release notes                                                         |
 
 ---
 
@@ -33,4 +33,4 @@ Browser (React 18 + Vite)  →  Express API (port 3001)  →  Gemini / Ollama
 
 ---
 
-*This page is auto-synced from `.github/wiki/Home.md` on every push to `main`.*
+_This page is auto-synced from `.github/wiki/Home.md` on every push to `main`._

CHANGELOG.md

@@ -7,10 +7,10 @@ All notable changes to this project will be documented in this file.
 ### Added
 
 - **Smarter pixel agents** — office canvas agents now react to four live signals instead of just test-run state:
-  1. *Chat-reactive*: `agentStatuses` ('working'/'waiting') is wired to `setAgentActive` so pixel agents walk to their desk and type whenever they are responding to a chat message, and wander again when idle (`PlaywrightDashboard.tsx`).
-  2. *Emotional reactions*: when a test run completes, all agents briefly show a green checkmark bubble (pass) or red ✗ bubble (fail) and celebrate by staying at their desks for 2.5 s (`officeState.ts`, `PlaywrightDashboard.tsx`).
-  3. *Handoff bubbles*: in multi-agent chat, the finishing agent gets a waiting bubble on `turn_done` — a visual "passing the baton" moment before the next agent activates (`PlaywrightDashboard.tsx`).
-  4. *New bubble types* — `react_pass` (light-green bg, bright checkmark) and `react_fail` (pink bg, red ×) pixel-art speech bubbles with auto-fade timer, matching the existing sprite pipeline (`types.ts`, `spriteData.ts`, `renderer.ts`).
+  1. _Chat-reactive_: `agentStatuses` ('working'/'waiting') is wired to `setAgentActive` so pixel agents walk to their desk and type whenever they are responding to a chat message, and wander again when idle (`PlaywrightDashboard.tsx`).
+  2. _Emotional reactions_: when a test run completes, all agents briefly show a green checkmark bubble (pass) or red ✗ bubble (fail) and celebrate by staying at their desks for 2.5 s (`officeState.ts`, `PlaywrightDashboard.tsx`).
+  3. _Handoff bubbles_: in multi-agent chat, the finishing agent gets a waiting bubble on `turn_done` — a visual "passing the baton" moment before the next agent activates (`PlaywrightDashboard.tsx`).
+  4. _New bubble types_ — `react_pass` (light-green bg, bright checkmark) and `react_fail` (pink bg, red ×) pixel-art speech bubbles with auto-fade timer, matching the existing sprite pipeline (`types.ts`, `spriteData.ts`, `renderer.ts`).
 - **SQLite run history** — replaces per-run JSON files in `test-results/runs/` with a
   single `better-sqlite3` database (`test-results/runs.db`). `archiveRun()` INSERTs into
   the DB; `/api/playwright/history`, `/api/playwright/results/:runId`, and
@@ -32,6 +32,16 @@ All notable changes to this project will be documented in this file.
   (`RUN_HISTORY_KEY`, seed-on-mount effect, write-on-change effect) to prevent stale
   pre-migration data surfacing on page load (`ui/src/pages/PlaywrightDashboard.tsx`).
 
+### Performance
+
+- **Fast Playwright startup** — replaced `npx playwright` with a direct
+  `node node_modules/@playwright/test/cli.js` spawn in both `/api/playwright/run`
+  and `/api/run-dynamic-test`. Eliminates `npx` + shell resolution overhead that
+  caused ~2-minute delays before tests began on Windows (`server/index.ts`).
+- **Non-blocking remote warmup** — warmup HTTP ping to Render/remote baseUrl now
+  fires concurrently with Playwright startup instead of blocking it; timeout
+  reduced from 55 s to 30 s (`server/index.ts`).
+
 ### Fixed
 
 - **Stop button race condition** — added `runAbortRef = useRef<AbortController | null>(null)`

README.md

@@ -31,7 +31,7 @@ Type a request in the chat. **Edi M**, the Team Manager, analyses your intent an
 
 |      | Feature                           | Description                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           |
 | ---- | --------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| 🏢   | **2D Pixel-Art Office**           | 7 specialist agents at animated desks — agents walk to their desk and type live while responding to chat, celebrate with checkmark bubbles on passing runs, show red ✗ bubbles on failures, and display handoff bubbles when passing work between agents in multi-turn conversations                                                                                                                                                                                                                                                   |
+| 🏢   | **2D Pixel-Art Office**           | 7 specialist agents at animated desks — agents walk to their desk and type live while responding to chat, celebrate with checkmark bubbles on passing runs, show red ✗ bubbles on failures, and display handoff bubbles when passing work between agents in multi-turn conversations                                                                                                                                                                                                                                                  |
 | 🤖   | **Edi M — Team Manager**          | Orchestrator agent that analyses your request and routes it to the right specialist                                                                                                                                                                                                                                                                                                                                                                                                                                                   |
 | 🤝   | **Multi-Agent Collaboration**     | Tag two or more specialists (or @Edi M) and they iterate: primary drafts → critic reviews → primary refines → manager delivers the polished synthesis. Each turn streams into its own chat bubble with a role pill (primary / critic / synthesis). Chat history and image attachments are forwarded to the initial primary turn so vision prompts and follow-up context are preserved across collaborative rounds                                                                                                                     |
 | ✨🦙 | **Gemini + Ollama**               | Switch between cloud Gemini and fully-local Ollama models with a single click — no restart required                                                                                                                                                                                                                                                                                                                                                                                                                                   |
@@ -70,7 +70,7 @@ Type a request in the chat. **Edi M**, the Team Manager, analyses your intent an
 │                             streams token chunks back via SSE        │
 │                                                                      │
 │  /api/run-dynamic-test                                               │
-│    Phase 1 → spawn  npx playwright test  →  stream stdout           │
+│    Phase 1 → spawn  node @playwright/test/cli.js  →  stream stdout  │
 │    Phase 2 → read   pw-results.json      →  Edi M AI summary        │
 │              classify failures  →  stream summary_chunk events       │
 └──────────────────────────────────────────────────────────────────────┘

server/index.ts

@@ -1212,7 +1212,7 @@ app.post('/api/playwright/run', async (req, res) => {
   };
   env.PW_RUNTIME_CONFIG = JSON.stringify(runtimeConfig);
 
-  send(`[INFO] Running: npx playwright ${args.join(' ')}`);
+  send(`[INFO] Running: playwright ${args.join(' ')}`);
   if (runtimeConfig.baseUrl) send(`[INFO] baseUrl  → ${runtimeConfig.baseUrl}`);
   if (runtimeConfig.timeout) send(`[INFO] timeout  → ${runtimeConfig.timeout}ms`);
   if (runtimeConfig.workers) send(`[INFO] workers  → ${runtimeConfig.workers}`);
@@ -1223,35 +1223,40 @@ app.post('/api/playwright/run', async (req, res) => {
   // Archive key: bare filenames joined (clean display in Run History)
   const archiveSpec = bareNames.length > 0 ? bareNames.join(',') : spec;
 
-  // ── Remote server warmup ──────────────────────────────────────────────────────
-  // Render free-tier sleeps after ~15 min of inactivity.  A single HTTP GET
-  // before spawning Playwright wakes the dyno so tests don't cold-start timeout.
+  // ── Remote server warmup (non-blocking) ──────────────────────────────────────
+  // Render free-tier sleeps after ~15 min of inactivity. Fire the warmup ping
+  // concurrently with Playwright startup so it doesn't add to wall-clock time.
   const targetUrl = (runtimeConfig.baseUrl as string | undefined) ?? '';
   if (targetUrl && !/localhost|127\.0\.0\.1/.test(targetUrl)) {
     send(`[INFO] Warming up ${targetUrl} …`);
-    try {
-      const httpMod = await import(targetUrl.startsWith('https') ? 'https' : 'http');
-      await new Promise<void>((resolve, reject) => {
-        const req = httpMod.default.get(targetUrl, (r: { resume: () => void }) => {
-          r.resume();
-          resolve();
-        });
-        req.setTimeout(55_000, () => {
-          req.destroy();
-          reject(new Error('warmup timeout'));
+    void (async () => {
+      try {
+        const httpMod = await import(targetUrl.startsWith('https') ? 'https' : 'http');
+        await new Promise<void>((resolve, reject) => {
+          const req = httpMod.default.get(targetUrl, (r: { resume: () => void }) => {
+            r.resume();
+            resolve();
+          });
+          req.setTimeout(30_000, () => {
+            req.destroy();
+            reject(new Error('warmup timeout'));
+          });
+          req.on('error', (e: Error) => reject(e));
         });
-        req.on('error', (e: Error) => reject(e));
-      });
-      send(`[INFO] Server is awake ✓`);
-    } catch (err: unknown) {
-      send(`[WARN] Warmup failed: ${String(err)} — tests will proceed but may be slow`);
-    }
+        send(`[INFO] Server is awake ✓`);
+      } catch (err: unknown) {
+        send(`[WARN] Warmup failed: ${String(err)} — tests may be slow`);
+      }
+    })();
   }
 
-  const child = spawn('npx', ['playwright', ...args], {
+  // Use the local playwright CLI directly — avoids npx resolution overhead
+  // (saves 30-90s on Windows where npx+shell startup is expensive).
+  const pwCli = path.join(root, 'node_modules', '@playwright', 'test', 'cli.js');
+  const child = spawn(process.execPath, [pwCli, ...args], {
     cwd: root,
     env,
-    shell: process.platform === 'win32',
+    shell: false,
   });
 
   // Capture log lines so they're (a) persisted with the archived run and
@@ -1571,7 +1576,7 @@ app.post('/api/run-dynamic-test', async (req, res) => {
   try {
     await fs.writeFile(tmpFile, authorLine + code, 'utf-8');
     sendStr(`[INFO] Dynamic spec written by ${agentName ?? 'agent'}`);
-    sendStr('[INFO] Running: npx playwright test _dynamic_agent_test.spec.ts');
+    sendStr('[INFO] Running: playwright test _dynamic_agent_test.spec.ts');
   } catch (err: unknown) {
     sendStr(`[ERROR] Failed to write test file: ${String(err)}`);
     sendStr('[DONE] Finished with exit code 1');
@@ -1585,10 +1590,11 @@ app.post('/api/run-dynamic-test', async (req, res) => {
     PW_RUNTIME_CONFIG: JSON.stringify({ testMatch: ['**/_dynamic_agent_test.spec.ts'] }),
   };
 
-  const child = spawn('npx', ['playwright', 'test'], {
+  const pwCli = path.join(root, 'node_modules', '@playwright', 'test', 'cli.js');
+  const child = spawn(process.execPath, [pwCli, 'test'], {
     cwd: root,
     env,
-    shell: process.platform === 'win32',
+    shell: false,
   });
 
   // Capture every log line so we can persist them with the archived run.